1 /*
2 * CDDL HEADER START
3 *
4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
7 *
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
12 *
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
18 *
19 * CDDL HEADER END
20 */
21 /*
22 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
23 * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
24 * Copyright (c) 2011 by Delphix. All rights reserved.
25 */
26
27 /*
28 * Pool import support functions.
29 *
30 * To import a pool, we rely on reading the configuration information from the
31 * ZFS label of each device. If we successfully read the label, then we
32 * organize the configuration information in the following hierarchy:
33 *
34 * pool guid -> toplevel vdev guid -> label txg
35 *
36 * Duplicate entries matching this same tuple will be discarded. Once we have
37 * examined every device, we pick the best label txg config for each toplevel
38 * vdev. We then arrange these toplevel vdevs into a complete pool config, and
39 * update any paths that have changed. Finally, we attempt to import the pool
40 * using our derived config, and record the results.
41 */
42
43 #include <ctype.h>
44 #include <devid.h>
45 #include <dirent.h>
46 #include <errno.h>
47 #include <libintl.h>
48 #include <stddef.h>
49 #include <stdlib.h>
50 #include <string.h>
51 #include <sys/stat.h>
52 #include <unistd.h>
53 #include <fcntl.h>
54 #include <sys/vtoc.h>
55 #include <sys/dktp/fdisk.h>
56 #include <sys/efi_partition.h>
57 #include <thread_pool.h>
58
59 #include <sys/vdev_impl.h>
60
61 #include "libzfs.h"
62 #include "libzfs_impl.h"
63
64 /*
65 * Intermediate structures used to gather configuration information.
66 */
67 typedef struct config_entry {
68 uint64_t ce_txg;
69 nvlist_t *ce_config;
70 struct config_entry *ce_next;
71 } config_entry_t;
72
73 typedef struct vdev_entry {
74 uint64_t ve_guid;
75 config_entry_t *ve_configs;
76 struct vdev_entry *ve_next;
77 } vdev_entry_t;
78
79 typedef struct pool_entry {
80 uint64_t pe_guid;
81 vdev_entry_t *pe_vdevs;
82 struct pool_entry *pe_next;
83 } pool_entry_t;
84
85 typedef struct name_entry {
86 char *ne_name;
87 uint64_t ne_guid;
88 struct name_entry *ne_next;
89 } name_entry_t;
90
91 typedef struct pool_list {
92 pool_entry_t *pools;
93 name_entry_t *names;
94 } pool_list_t;
95
96 static char *
97 get_devid(const char *path)
98 {
99 int fd;
100 ddi_devid_t devid;
101 char *minor, *ret;
102
103 if ((fd = open(path, O_RDONLY)) < 0)
104 return (NULL);
105
106 minor = NULL;
107 ret = NULL;
108 if (devid_get(fd, &devid) == 0) {
109 if (devid_get_minor_name(fd, &minor) == 0)
110 ret = devid_str_encode(devid, minor);
111 if (minor != NULL)
112 devid_str_free(minor);
113 devid_free(devid);
114 }
115 (void) close(fd);
116
117 return (ret);
118 }
119
120
121 /*
122 * Go through and fix up any path and/or devid information for the given vdev
123 * configuration.
124 */
125 static int
126 fix_paths(nvlist_t *nv, name_entry_t *names)
127 {
128 nvlist_t **child;
129 uint_t c, children;
130 uint64_t guid;
131 name_entry_t *ne, *best;
132 char *path, *devid;
133 int matched;
134
135 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
136 &child, &children) == 0) {
137 for (c = 0; c < children; c++)
138 if (fix_paths(child[c], names) != 0)
139 return (-1);
140 return (0);
141 }
142
143 /*
144 * This is a leaf (file or disk) vdev. In either case, go through
145 * the name list and see if we find a matching guid. If so, replace
146 * the path and see if we can calculate a new devid.
147 *
148 * There may be multiple names associated with a particular guid, in
149 * which case we have overlapping slices or multiple paths to the same
150 * disk. If this is the case, then we want to pick the path that is
151 * the most similar to the original, where "most similar" is the number
152 * of matching characters starting from the end of the path. This will
153 * preserve slice numbers even if the disks have been reorganized, and
154 * will also catch preferred disk names if multiple paths exist.
155 */
156 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &guid) == 0);
157 if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) != 0)
158 path = NULL;
159
160 matched = 0;
161 best = NULL;
162 for (ne = names; ne != NULL; ne = ne->ne_next) {
163 if (ne->ne_guid == guid) {
164 const char *src, *dst;
165 int count;
166
167 if (path == NULL) {
168 best = ne;
169 break;
170 }
171
172 src = ne->ne_name + strlen(ne->ne_name) - 1;
173 dst = path + strlen(path) - 1;
174 for (count = 0; src >= ne->ne_name && dst >= path;
175 src--, dst--, count++)
176 if (*src != *dst)
177 break;
178
179 /*
180 * At this point, 'count' is the number of characters
181 * matched from the end.
182 */
183 if (count > matched || best == NULL) {
184 best = ne;
185 matched = count;
186 }
187 }
188 }
189
190 if (best == NULL)
191 return (0);
192
193 if (nvlist_add_string(nv, ZPOOL_CONFIG_PATH, best->ne_name) != 0)
194 return (-1);
195
196 if ((devid = get_devid(best->ne_name)) == NULL) {
197 (void) nvlist_remove_all(nv, ZPOOL_CONFIG_DEVID);
198 } else {
199 if (nvlist_add_string(nv, ZPOOL_CONFIG_DEVID, devid) != 0)
200 return (-1);
201 devid_str_free(devid);
202 }
203
204 return (0);
205 }
206
207 /*
208 * Add the given configuration to the list of known devices.
209 */
210 static int
211 add_config(libzfs_handle_t *hdl, pool_list_t *pl, const char *path,
212 nvlist_t *config)
213 {
214 uint64_t pool_guid, vdev_guid, top_guid, txg, state;
215 pool_entry_t *pe;
216 vdev_entry_t *ve;
217 config_entry_t *ce;
218 name_entry_t *ne;
219
220 /*
221 * If this is a hot spare not currently in use or level 2 cache
222 * device, add it to the list of names to translate, but don't do
223 * anything else.
224 */
225 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE,
226 &state) == 0 &&
227 (state == POOL_STATE_SPARE || state == POOL_STATE_L2CACHE) &&
228 nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID, &vdev_guid) == 0) {
229 if ((ne = zfs_alloc(hdl, sizeof (name_entry_t))) == NULL)
230 return (-1);
231
232 if ((ne->ne_name = zfs_strdup(hdl, path)) == NULL) {
233 free(ne);
234 return (-1);
235 }
236 ne->ne_guid = vdev_guid;
237 ne->ne_next = pl->names;
238 pl->names = ne;
239 return (0);
240 }
241
242 /*
243 * If we have a valid config but cannot read any of these fields, then
244 * it means we have a half-initialized label. In vdev_label_init()
245 * we write a label with txg == 0 so that we can identify the device
246 * in case the user refers to the same disk later on. If we fail to
247 * create the pool, we'll be left with a label in this state
248 * which should not be considered part of a valid pool.
249 */
250 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
251 &pool_guid) != 0 ||
252 nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID,
253 &vdev_guid) != 0 ||
254 nvlist_lookup_uint64(config, ZPOOL_CONFIG_TOP_GUID,
255 &top_guid) != 0 ||
256 nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_TXG,
257 &txg) != 0 || txg == 0) {
258 nvlist_free(config);
259 return (0);
260 }
261
262 /*
263 * First, see if we know about this pool. If not, then add it to the
264 * list of known pools.
265 */
266 for (pe = pl->pools; pe != NULL; pe = pe->pe_next) {
267 if (pe->pe_guid == pool_guid)
268 break;
269 }
270
271 if (pe == NULL) {
272 if ((pe = zfs_alloc(hdl, sizeof (pool_entry_t))) == NULL) {
273 nvlist_free(config);
274 return (-1);
275 }
276 pe->pe_guid = pool_guid;
277 pe->pe_next = pl->pools;
278 pl->pools = pe;
279 }
280
281 /*
282 * Second, see if we know about this toplevel vdev. Add it if its
283 * missing.
284 */
285 for (ve = pe->pe_vdevs; ve != NULL; ve = ve->ve_next) {
286 if (ve->ve_guid == top_guid)
287 break;
288 }
289
290 if (ve == NULL) {
291 if ((ve = zfs_alloc(hdl, sizeof (vdev_entry_t))) == NULL) {
292 nvlist_free(config);
293 return (-1);
294 }
295 ve->ve_guid = top_guid;
296 ve->ve_next = pe->pe_vdevs;
297 pe->pe_vdevs = ve;
298 }
299
300 /*
301 * Third, see if we have a config with a matching transaction group. If
302 * so, then we do nothing. Otherwise, add it to the list of known
303 * configs.
304 */
305 for (ce = ve->ve_configs; ce != NULL; ce = ce->ce_next) {
306 if (ce->ce_txg == txg)
307 break;
308 }
309
310 if (ce == NULL) {
311 if ((ce = zfs_alloc(hdl, sizeof (config_entry_t))) == NULL) {
312 nvlist_free(config);
313 return (-1);
314 }
315 ce->ce_txg = txg;
316 ce->ce_config = config;
317 ce->ce_next = ve->ve_configs;
318 ve->ve_configs = ce;
319 } else {
320 nvlist_free(config);
321 }
322
323 /*
324 * At this point we've successfully added our config to the list of
325 * known configs. The last thing to do is add the vdev guid -> path
326 * mappings so that we can fix up the configuration as necessary before
327 * doing the import.
328 */
329 if ((ne = zfs_alloc(hdl, sizeof (name_entry_t))) == NULL)
330 return (-1);
331
332 if ((ne->ne_name = zfs_strdup(hdl, path)) == NULL) {
333 free(ne);
334 return (-1);
335 }
336
337 ne->ne_guid = vdev_guid;
338 ne->ne_next = pl->names;
339 pl->names = ne;
340
341 return (0);
342 }
343
344 /*
345 * Returns true if the named pool matches the given GUID.
346 */
347 static int
348 pool_active(libzfs_handle_t *hdl, const char *name, uint64_t guid,
349 boolean_t *isactive)
350 {
351 zpool_handle_t *zhp;
352 uint64_t theguid;
353
354 if (zpool_open_silent(hdl, name, &zhp) != 0)
355 return (-1);
356
357 if (zhp == NULL) {
358 *isactive = B_FALSE;
359 return (0);
360 }
361
362 verify(nvlist_lookup_uint64(zhp->zpool_config, ZPOOL_CONFIG_POOL_GUID,
363 &theguid) == 0);
364
365 zpool_close(zhp);
366
367 *isactive = (theguid == guid);
368 return (0);
369 }
370
371 static nvlist_t *
372 refresh_config(libzfs_handle_t *hdl, nvlist_t *config)
373 {
374 nvlist_t *nvl;
375 zfs_cmd_t zc = { 0 };
376 int err;
377
378 if (zcmd_write_conf_nvlist(hdl, &zc, config) != 0)
379 return (NULL);
380
381 if (zcmd_alloc_dst_nvlist(hdl, &zc,
382 zc.zc_nvlist_conf_size * 2) != 0) {
383 zcmd_free_nvlists(&zc);
384 return (NULL);
385 }
386
387 while ((err = ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_TRYIMPORT,
388 &zc)) != 0 && errno == ENOMEM) {
389 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
390 zcmd_free_nvlists(&zc);
391 return (NULL);
392 }
393 }
394
395 if (err) {
396 zcmd_free_nvlists(&zc);
397 return (NULL);
398 }
399
400 if (zcmd_read_dst_nvlist(hdl, &zc, &nvl) != 0) {
401 zcmd_free_nvlists(&zc);
402 return (NULL);
403 }
404
405 zcmd_free_nvlists(&zc);
406 return (nvl);
407 }
408
409 /*
410 * Determine if the vdev id is a hole in the namespace.
411 */
412 boolean_t
413 vdev_is_hole(uint64_t *hole_array, uint_t holes, uint_t id)
414 {
415 for (int c = 0; c < holes; c++) {
416
417 /* Top-level is a hole */
418 if (hole_array[c] == id)
419 return (B_TRUE);
420 }
421 return (B_FALSE);
422 }
423
424 /*
425 * Convert our list of pools into the definitive set of configurations. We
426 * start by picking the best config for each toplevel vdev. Once that's done,
427 * we assemble the toplevel vdevs into a full config for the pool. We make a
428 * pass to fix up any incorrect paths, and then add it to the main list to
429 * return to the user.
430 */
431 static nvlist_t *
432 get_configs(libzfs_handle_t *hdl, pool_list_t *pl, boolean_t active_ok)
433 {
434 pool_entry_t *pe;
435 vdev_entry_t *ve;
436 config_entry_t *ce;
437 nvlist_t *ret = NULL, *config = NULL, *tmp, *nvtop, *nvroot;
438 nvlist_t **spares, **l2cache;
439 uint_t i, nspares, nl2cache;
440 boolean_t config_seen;
441 uint64_t best_txg;
442 char *name, *hostname, *comment;
443 uint64_t version, guid;
444 uint_t children = 0;
445 nvlist_t **child = NULL;
446 uint_t holes;
447 uint64_t *hole_array, max_id;
448 uint_t c;
449 boolean_t isactive;
450 uint64_t hostid;
451 nvlist_t *nvl;
452 boolean_t found_one = B_FALSE;
453 boolean_t valid_top_config = B_FALSE;
454
455 if (nvlist_alloc(&ret, 0, 0) != 0)
456 goto nomem;
457
458 for (pe = pl->pools; pe != NULL; pe = pe->pe_next) {
459 uint64_t id, max_txg = 0;
460
461 if (nvlist_alloc(&config, NV_UNIQUE_NAME, 0) != 0)
462 goto nomem;
463 config_seen = B_FALSE;
464
465 /*
466 * Iterate over all toplevel vdevs. Grab the pool configuration
467 * from the first one we find, and then go through the rest and
468 * add them as necessary to the 'vdevs' member of the config.
469 */
470 for (ve = pe->pe_vdevs; ve != NULL; ve = ve->ve_next) {
471
472 /*
473 * Determine the best configuration for this vdev by
474 * selecting the config with the latest transaction
475 * group.
476 */
477 best_txg = 0;
478 for (ce = ve->ve_configs; ce != NULL;
479 ce = ce->ce_next) {
480
481 if (ce->ce_txg > best_txg) {
482 tmp = ce->ce_config;
483 best_txg = ce->ce_txg;
484 }
485 }
486
487 /*
488 * We rely on the fact that the max txg for the
489 * pool will contain the most up-to-date information
490 * about the valid top-levels in the vdev namespace.
491 */
492 if (best_txg > max_txg) {
493 (void) nvlist_remove(config,
494 ZPOOL_CONFIG_VDEV_CHILDREN,
495 DATA_TYPE_UINT64);
496 (void) nvlist_remove(config,
497 ZPOOL_CONFIG_HOLE_ARRAY,
498 DATA_TYPE_UINT64_ARRAY);
499
500 max_txg = best_txg;
501 hole_array = NULL;
502 holes = 0;
503 max_id = 0;
504 valid_top_config = B_FALSE;
505
506 if (nvlist_lookup_uint64(tmp,
507 ZPOOL_CONFIG_VDEV_CHILDREN, &max_id) == 0) {
508 verify(nvlist_add_uint64(config,
509 ZPOOL_CONFIG_VDEV_CHILDREN,
510 max_id) == 0);
511 valid_top_config = B_TRUE;
512 }
513
514 if (nvlist_lookup_uint64_array(tmp,
515 ZPOOL_CONFIG_HOLE_ARRAY, &hole_array,
516 &holes) == 0) {
517 verify(nvlist_add_uint64_array(config,
518 ZPOOL_CONFIG_HOLE_ARRAY,
519 hole_array, holes) == 0);
520 }
521 }
522
523 if (!config_seen) {
524 /*
525 * Copy the relevant pieces of data to the pool
526 * configuration:
527 *
528 * version
529 * pool guid
530 * name
531 * comment (if available)
532 * pool state
533 * hostid (if available)
534 * hostname (if available)
535 */
536 uint64_t state;
537
538 verify(nvlist_lookup_uint64(tmp,
539 ZPOOL_CONFIG_VERSION, &version) == 0);
540 if (nvlist_add_uint64(config,
541 ZPOOL_CONFIG_VERSION, version) != 0)
542 goto nomem;
543 verify(nvlist_lookup_uint64(tmp,
544 ZPOOL_CONFIG_POOL_GUID, &guid) == 0);
545 if (nvlist_add_uint64(config,
546 ZPOOL_CONFIG_POOL_GUID, guid) != 0)
547 goto nomem;
548 verify(nvlist_lookup_string(tmp,
549 ZPOOL_CONFIG_POOL_NAME, &name) == 0);
550 if (nvlist_add_string(config,
551 ZPOOL_CONFIG_POOL_NAME, name) != 0)
552 goto nomem;
553
554 /*
555 * COMMENT is optional, don't bail if it's not
556 * there, instead, set it to NULL.
557 */
558 if (nvlist_lookup_string(tmp,
559 ZPOOL_CONFIG_COMMENT, &comment) != 0)
560 comment = NULL;
561 else if (nvlist_add_string(config,
562 ZPOOL_CONFIG_COMMENT, comment) != 0)
563 goto nomem;
564
565 verify(nvlist_lookup_uint64(tmp,
566 ZPOOL_CONFIG_POOL_STATE, &state) == 0);
567 if (nvlist_add_uint64(config,
568 ZPOOL_CONFIG_POOL_STATE, state) != 0)
569 goto nomem;
570
571 hostid = 0;
572 if (nvlist_lookup_uint64(tmp,
573 ZPOOL_CONFIG_HOSTID, &hostid) == 0) {
574 if (nvlist_add_uint64(config,
575 ZPOOL_CONFIG_HOSTID, hostid) != 0)
576 goto nomem;
577 verify(nvlist_lookup_string(tmp,
578 ZPOOL_CONFIG_HOSTNAME,
579 &hostname) == 0);
580 if (nvlist_add_string(config,
581 ZPOOL_CONFIG_HOSTNAME,
582 hostname) != 0)
583 goto nomem;
584 }
585
586 config_seen = B_TRUE;
587 }
588
589 /*
590 * Add this top-level vdev to the child array.
591 */
592 verify(nvlist_lookup_nvlist(tmp,
593 ZPOOL_CONFIG_VDEV_TREE, &nvtop) == 0);
594 verify(nvlist_lookup_uint64(nvtop, ZPOOL_CONFIG_ID,
595 &id) == 0);
596
597 if (id >= children) {
598 nvlist_t **newchild;
599
600 newchild = zfs_alloc(hdl, (id + 1) *
601 sizeof (nvlist_t *));
602 if (newchild == NULL)
603 goto nomem;
604
605 for (c = 0; c < children; c++)
606 newchild[c] = child[c];
607
608 free(child);
609 child = newchild;
610 children = id + 1;
611 }
612 if (nvlist_dup(nvtop, &child[id], 0) != 0)
613 goto nomem;
614
615 }
616
617 /*
618 * If we have information about all the top-levels then
619 * clean up the nvlist which we've constructed. This
620 * means removing any extraneous devices that are
621 * beyond the valid range or adding devices to the end
622 * of our array which appear to be missing.
623 */
624 if (valid_top_config) {
625 if (max_id < children) {
626 for (c = max_id; c < children; c++)
627 nvlist_free(child[c]);
628 children = max_id;
629 } else if (max_id > children) {
630 nvlist_t **newchild;
631
632 newchild = zfs_alloc(hdl, (max_id) *
633 sizeof (nvlist_t *));
634 if (newchild == NULL)
635 goto nomem;
636
637 for (c = 0; c < children; c++)
638 newchild[c] = child[c];
639
640 free(child);
641 child = newchild;
642 children = max_id;
643 }
644 }
645
646 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
647 &guid) == 0);
648
649 /*
650 * The vdev namespace may contain holes as a result of
651 * device removal. We must add them back into the vdev
652 * tree before we process any missing devices.
653 */
654 if (holes > 0) {
655 ASSERT(valid_top_config);
656
657 for (c = 0; c < children; c++) {
658 nvlist_t *holey;
659
660 if (child[c] != NULL ||
661 !vdev_is_hole(hole_array, holes, c))
662 continue;
663
664 if (nvlist_alloc(&holey, NV_UNIQUE_NAME,
665 0) != 0)
666 goto nomem;
667
668 /*
669 * Holes in the namespace are treated as
670 * "hole" top-level vdevs and have a
671 * special flag set on them.
672 */
673 if (nvlist_add_string(holey,
674 ZPOOL_CONFIG_TYPE,
675 VDEV_TYPE_HOLE) != 0 ||
676 nvlist_add_uint64(holey,
677 ZPOOL_CONFIG_ID, c) != 0 ||
678 nvlist_add_uint64(holey,
679 ZPOOL_CONFIG_GUID, 0ULL) != 0)
680 goto nomem;
681 child[c] = holey;
682 }
683 }
684
685 /*
686 * Look for any missing top-level vdevs. If this is the case,
687 * create a faked up 'missing' vdev as a placeholder. We cannot
688 * simply compress the child array, because the kernel performs
689 * certain checks to make sure the vdev IDs match their location
690 * in the configuration.
691 */
692 for (c = 0; c < children; c++) {
693 if (child[c] == NULL) {
694 nvlist_t *missing;
695 if (nvlist_alloc(&missing, NV_UNIQUE_NAME,
696 0) != 0)
697 goto nomem;
698 if (nvlist_add_string(missing,
699 ZPOOL_CONFIG_TYPE,
700 VDEV_TYPE_MISSING) != 0 ||
701 nvlist_add_uint64(missing,
702 ZPOOL_CONFIG_ID, c) != 0 ||
703 nvlist_add_uint64(missing,
704 ZPOOL_CONFIG_GUID, 0ULL) != 0) {
705 nvlist_free(missing);
706 goto nomem;
707 }
708 child[c] = missing;
709 }
710 }
711
712 /*
713 * Put all of this pool's top-level vdevs into a root vdev.
714 */
715 if (nvlist_alloc(&nvroot, NV_UNIQUE_NAME, 0) != 0)
716 goto nomem;
717 if (nvlist_add_string(nvroot, ZPOOL_CONFIG_TYPE,
718 VDEV_TYPE_ROOT) != 0 ||
719 nvlist_add_uint64(nvroot, ZPOOL_CONFIG_ID, 0ULL) != 0 ||
720 nvlist_add_uint64(nvroot, ZPOOL_CONFIG_GUID, guid) != 0 ||
721 nvlist_add_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
722 child, children) != 0) {
723 nvlist_free(nvroot);
724 goto nomem;
725 }
726
727 for (c = 0; c < children; c++)
728 nvlist_free(child[c]);
729 free(child);
730 children = 0;
731 child = NULL;
732
733 /*
734 * Go through and fix up any paths and/or devids based on our
735 * known list of vdev GUID -> path mappings.
736 */
737 if (fix_paths(nvroot, pl->names) != 0) {
738 nvlist_free(nvroot);
739 goto nomem;
740 }
741
742 /*
743 * Add the root vdev to this pool's configuration.
744 */
745 if (nvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
746 nvroot) != 0) {
747 nvlist_free(nvroot);
748 goto nomem;
749 }
750 nvlist_free(nvroot);
751
752 /*
753 * zdb uses this path to report on active pools that were
754 * imported or created using -R.
755 */
756 if (active_ok)
757 goto add_pool;
758
759 /*
760 * Determine if this pool is currently active, in which case we
761 * can't actually import it.
762 */
763 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
764 &name) == 0);
765 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
766 &guid) == 0);
767
768 if (pool_active(hdl, name, guid, &isactive) != 0)
769 goto error;
770
771 if (isactive) {
772 nvlist_free(config);
773 config = NULL;
774 continue;
775 }
776
777 if ((nvl = refresh_config(hdl, config)) == NULL) {
778 nvlist_free(config);
779 config = NULL;
780 continue;
781 }
782
783 nvlist_free(config);
784 config = nvl;
785
786 /*
787 * Go through and update the paths for spares, now that we have
788 * them.
789 */
790 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
791 &nvroot) == 0);
792 if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
793 &spares, &nspares) == 0) {
794 for (i = 0; i < nspares; i++) {
795 if (fix_paths(spares[i], pl->names) != 0)
796 goto nomem;
797 }
798 }
799
800 /*
801 * Update the paths for l2cache devices.
802 */
803 if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
804 &l2cache, &nl2cache) == 0) {
805 for (i = 0; i < nl2cache; i++) {
806 if (fix_paths(l2cache[i], pl->names) != 0)
807 goto nomem;
808 }
809 }
810
811 /*
812 * Restore the original information read from the actual label.
813 */
814 (void) nvlist_remove(config, ZPOOL_CONFIG_HOSTID,
815 DATA_TYPE_UINT64);
816 (void) nvlist_remove(config, ZPOOL_CONFIG_HOSTNAME,
817 DATA_TYPE_STRING);
818 if (hostid != 0) {
819 verify(nvlist_add_uint64(config, ZPOOL_CONFIG_HOSTID,
820 hostid) == 0);
821 verify(nvlist_add_string(config, ZPOOL_CONFIG_HOSTNAME,
822 hostname) == 0);
823 }
824
825 add_pool:
826 /*
827 * Add this pool to the list of configs.
828 */
829 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
830 &name) == 0);
831 if (nvlist_add_nvlist(ret, name, config) != 0)
832 goto nomem;
833
834 found_one = B_TRUE;
835 nvlist_free(config);
836 config = NULL;
837 }
838
839 if (!found_one) {
840 nvlist_free(ret);
841 ret = NULL;
842 }
843
844 return (ret);
845
846 nomem:
847 (void) no_memory(hdl);
848 error:
849 nvlist_free(config);
850 nvlist_free(ret);
851 for (c = 0; c < children; c++)
852 nvlist_free(child[c]);
853 free(child);
854
855 return (NULL);
856 }
857
858 /*
859 * Return the offset of the given label.
860 */
861 static uint64_t
862 label_offset(uint64_t size, int l)
863 {
864 ASSERT(P2PHASE_TYPED(size, sizeof (vdev_label_t), uint64_t) == 0);
865 return (l * sizeof (vdev_label_t) + (l < VDEV_LABELS / 2 ?
866 0 : size - VDEV_LABELS * sizeof (vdev_label_t)));
867 }
868
869 /*
870 * Given a file descriptor, read the label information and return an nvlist
871 * describing the configuration, if there is one.
872 */
873 int
874 zpool_read_label(int fd, nvlist_t **config)
875 {
876 struct stat64 statbuf;
877 int l;
878 vdev_label_t *label;
879 uint64_t state, txg, size;
880
881 *config = NULL;
882
883 if (fstat64(fd, &statbuf) == -1)
884 return (0);
885 size = P2ALIGN_TYPED(statbuf.st_size, sizeof (vdev_label_t), uint64_t);
886
887 if ((label = malloc(sizeof (vdev_label_t))) == NULL)
888 return (-1);
889
890 for (l = 0; l < VDEV_LABELS; l++) {
891 if (pread64(fd, label, sizeof (vdev_label_t),
892 label_offset(size, l)) != sizeof (vdev_label_t))
893 continue;
894
895 if (nvlist_unpack(label->vl_vdev_phys.vp_nvlist,
896 sizeof (label->vl_vdev_phys.vp_nvlist), config, 0) != 0)
897 continue;
898
899 if (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_STATE,
900 &state) != 0 || state > POOL_STATE_L2CACHE) {
901 nvlist_free(*config);
902 continue;
903 }
904
905 if (state != POOL_STATE_SPARE && state != POOL_STATE_L2CACHE &&
906 (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_TXG,
907 &txg) != 0 || txg == 0)) {
908 nvlist_free(*config);
909 continue;
910 }
911
912 free(label);
913 return (0);
914 }
915
916 free(label);
917 *config = NULL;
918 return (0);
919 }
920
921 typedef struct rdsk_node {
922 char *rn_name;
923 int rn_dfd;
924 libzfs_handle_t *rn_hdl;
925 nvlist_t *rn_config;
926 avl_tree_t *rn_avl;
927 avl_node_t rn_node;
928 boolean_t rn_nozpool;
929 } rdsk_node_t;
930
931 static int
932 slice_cache_compare(const void *arg1, const void *arg2)
933 {
934 const char *nm1 = ((rdsk_node_t *)arg1)->rn_name;
935 const char *nm2 = ((rdsk_node_t *)arg2)->rn_name;
936 char *nm1slice, *nm2slice;
937 int rv;
938
939 /*
940 * slices zero and two are the most likely to provide results,
941 * so put those first
942 */
943 nm1slice = strstr(nm1, "s0");
944 nm2slice = strstr(nm2, "s0");
945 if (nm1slice && !nm2slice) {
946 return (-1);
947 }
948 if (!nm1slice && nm2slice) {
949 return (1);
950 }
951 nm1slice = strstr(nm1, "s2");
952 nm2slice = strstr(nm2, "s2");
953 if (nm1slice && !nm2slice) {
954 return (-1);
955 }
956 if (!nm1slice && nm2slice) {
957 return (1);
958 }
959
960 rv = strcmp(nm1, nm2);
961 if (rv == 0)
962 return (0);
963 return (rv > 0 ? 1 : -1);
964 }
965
966 static void
967 check_one_slice(avl_tree_t *r, char *diskname, uint_t partno,
968 diskaddr_t size, uint_t blksz)
969 {
970 rdsk_node_t tmpnode;
971 rdsk_node_t *node;
972 char sname[MAXNAMELEN];
973
974 tmpnode.rn_name = &sname[0];
975 (void) snprintf(tmpnode.rn_name, MAXNAMELEN, "%s%u",
976 diskname, partno);
977 /*
978 * protect against division by zero for disk labels that
979 * contain a bogus sector size
980 */
981 if (blksz == 0)
982 blksz = DEV_BSIZE;
983 /* too small to contain a zpool? */
984 if ((size < (SPA_MINDEVSIZE / blksz)) &&
985 (node = avl_find(r, &tmpnode, NULL)))
986 node->rn_nozpool = B_TRUE;
987 }
988
989 static void
990 nozpool_all_slices(avl_tree_t *r, const char *sname)
991 {
992 char diskname[MAXNAMELEN];
993 char *ptr;
994 int i;
995
996 (void) strncpy(diskname, sname, MAXNAMELEN);
997 if (((ptr = strrchr(diskname, 's')) == NULL) &&
998 ((ptr = strrchr(diskname, 'p')) == NULL))
999 return;
1000 ptr[0] = 's';
1001 ptr[1] = '\0';
1002 for (i = 0; i < NDKMAP; i++)
1003 check_one_slice(r, diskname, i, 0, 1);
1004 ptr[0] = 'p';
1005 for (i = 0; i <= FD_NUMPART; i++)
1006 check_one_slice(r, diskname, i, 0, 1);
1007 }
1008
1009 static void
1010 check_slices(avl_tree_t *r, int fd, const char *sname)
1011 {
1012 struct extvtoc vtoc;
1013 struct dk_gpt *gpt;
1014 char diskname[MAXNAMELEN];
1015 char *ptr;
1016 int i;
1017
1018 (void) strncpy(diskname, sname, MAXNAMELEN);
1019 if ((ptr = strrchr(diskname, 's')) == NULL || !isdigit(ptr[1]))
1020 return;
1021 ptr[1] = '\0';
1022
1023 if (read_extvtoc(fd, &vtoc) >= 0) {
1024 for (i = 0; i < NDKMAP; i++)
1025 check_one_slice(r, diskname, i,
1026 vtoc.v_part[i].p_size, vtoc.v_sectorsz);
1027 } else if (efi_alloc_and_read(fd, &gpt) >= 0) {
1028 /*
1029 * on x86 we'll still have leftover links that point
1030 * to slices s[9-15], so use NDKMAP instead
1031 */
1032 for (i = 0; i < NDKMAP; i++)
1033 check_one_slice(r, diskname, i,
1034 gpt->efi_parts[i].p_size, gpt->efi_lbasize);
1035 /* nodes p[1-4] are never used with EFI labels */
1036 ptr[0] = 'p';
1037 for (i = 1; i <= FD_NUMPART; i++)
1038 check_one_slice(r, diskname, i, 0, 1);
1039 efi_free(gpt);
1040 }
1041 }
1042
1043 static void
1044 zpool_open_func(void *arg)
1045 {
1046 rdsk_node_t *rn = arg;
1047 struct stat64 statbuf;
1048 nvlist_t *config;
1049 int fd;
1050
1051 if (rn->rn_nozpool)
1052 return;
1053 if ((fd = openat64(rn->rn_dfd, rn->rn_name, O_RDONLY)) < 0) {
1054 /* symlink to a device that's no longer there */
1055 if (errno == ENOENT)
1056 nozpool_all_slices(rn->rn_avl, rn->rn_name);
1057 return;
1058 }
1059 /*
1060 * Ignore failed stats. We only want regular
1061 * files, character devs and block devs.
1062 */
1063 if (fstat64(fd, &statbuf) != 0 ||
1064 (!S_ISREG(statbuf.st_mode) &&
1065 !S_ISCHR(statbuf.st_mode) &&
1066 !S_ISBLK(statbuf.st_mode))) {
1067 (void) close(fd);
1068 return;
1069 }
1070 /* this file is too small to hold a zpool */
1071 if (S_ISREG(statbuf.st_mode) &&
1072 statbuf.st_size < SPA_MINDEVSIZE) {
1073 (void) close(fd);
1074 return;
1075 } else if (!S_ISREG(statbuf.st_mode)) {
1076 /*
1077 * Try to read the disk label first so we don't have to
1078 * open a bunch of minor nodes that can't have a zpool.
1079 */
1080 check_slices(rn->rn_avl, fd, rn->rn_name);
1081 }
1082
1083 if ((zpool_read_label(fd, &config)) != 0) {
1084 (void) close(fd);
1085 (void) no_memory(rn->rn_hdl);
1086 return;
1087 }
1088 (void) close(fd);
1089
1090
1091 rn->rn_config = config;
1092 if (config != NULL) {
1093 assert(rn->rn_nozpool == B_FALSE);
1094 }
1095 }
1096
1097 /*
1098 * Given a file descriptor, clear (zero) the label information. This function
1099 * is currently only used in the appliance stack as part of the ZFS sysevent
1100 * module.
1101 */
1102 int
1103 zpool_clear_label(int fd)
1104 {
1105 struct stat64 statbuf;
1106 int l;
1107 vdev_label_t *label;
1108 uint64_t size;
1109
1110 if (fstat64(fd, &statbuf) == -1)
1111 return (0);
1112 size = P2ALIGN_TYPED(statbuf.st_size, sizeof (vdev_label_t), uint64_t);
1113
1114 if ((label = calloc(sizeof (vdev_label_t), 1)) == NULL)
1115 return (-1);
1116
1117 for (l = 0; l < VDEV_LABELS; l++) {
1118 if (pwrite64(fd, label, sizeof (vdev_label_t),
1119 label_offset(size, l)) != sizeof (vdev_label_t))
1120 return (-1);
1121 }
1122
1123 free(label);
1124 return (0);
1125 }
1126
1127 /*
1128 * Given a list of directories to search, find all pools stored on disk. This
1129 * includes partial pools which are not available to import. If no args are
1130 * given (argc is 0), then the default directory (/dev/dsk) is searched.
1131 * poolname or guid (but not both) are provided by the caller when trying
1132 * to import a specific pool.
1133 */
1134 static nvlist_t *
1135 zpool_find_import_impl(libzfs_handle_t *hdl, importargs_t *iarg)
1136 {
1137 int i, dirs = iarg->paths;
1138 DIR *dirp = NULL;
1139 struct dirent64 *dp;
1140 char path[MAXPATHLEN];
1141 char *end, **dir = iarg->path;
1142 size_t pathleft;
1143 nvlist_t *ret = NULL;
1144 static char *default_dir = "/dev/dsk";
1145 pool_list_t pools = { 0 };
1146 pool_entry_t *pe, *penext;
1147 vdev_entry_t *ve, *venext;
1148 config_entry_t *ce, *cenext;
1149 name_entry_t *ne, *nenext;
1150 avl_tree_t slice_cache;
1151 rdsk_node_t *slice;
1152 void *cookie;
1153
1154 if (dirs == 0) {
1155 dirs = 1;
1156 dir = &default_dir;
1157 }
1158
1159 /*
1160 * Go through and read the label configuration information from every
1161 * possible device, organizing the information according to pool GUID
1162 * and toplevel GUID.
1163 */
1164 for (i = 0; i < dirs; i++) {
1165 tpool_t *t;
1166 char *rdsk;
1167 int dfd;
1168
1169 /* use realpath to normalize the path */
1170 if (realpath(dir[i], path) == 0) {
1171 (void) zfs_error_fmt(hdl, EZFS_BADPATH,
1172 dgettext(TEXT_DOMAIN, "cannot open '%s'"), dir[i]);
1173 goto error;
1174 }
1175 end = &path[strlen(path)];
1176 *end++ = '/';
1177 *end = 0;
1178 pathleft = &path[sizeof (path)] - end;
1179
1180 /*
1181 * Using raw devices instead of block devices when we're
1182 * reading the labels skips a bunch of slow operations during
1183 * close(2) processing, so we replace /dev/dsk with /dev/rdsk.
1184 */
1185 if (strcmp(path, "/dev/dsk/") == 0)
1186 rdsk = "/dev/rdsk/";
1187 else
1188 rdsk = path;
1189
1190 if ((dfd = open64(rdsk, O_RDONLY)) < 0 ||
1191 (dirp = fdopendir(dfd)) == NULL) {
1192 zfs_error_aux(hdl, strerror(errno));
1193 (void) zfs_error_fmt(hdl, EZFS_BADPATH,
1194 dgettext(TEXT_DOMAIN, "cannot open '%s'"),
1195 rdsk);
1196 goto error;
1197 }
1198
1199 avl_create(&slice_cache, slice_cache_compare,
1200 sizeof (rdsk_node_t), offsetof(rdsk_node_t, rn_node));
1201 /*
1202 * This is not MT-safe, but we have no MT consumers of libzfs
1203 */
1204 while ((dp = readdir64(dirp)) != NULL) {
1205 const char *name = dp->d_name;
1206 if (name[0] == '.' &&
1207 (name[1] == 0 || (name[1] == '.' && name[2] == 0)))
1208 continue;
1209
1210 slice = zfs_alloc(hdl, sizeof (rdsk_node_t));
1211 slice->rn_name = zfs_strdup(hdl, name);
1212 slice->rn_avl = &slice_cache;
1213 slice->rn_dfd = dfd;
1214 slice->rn_hdl = hdl;
1215 slice->rn_nozpool = B_FALSE;
1216 avl_add(&slice_cache, slice);
1217 }
1218 /*
1219 * create a thread pool to do all of this in parallel;
1220 * rn_nozpool is not protected, so this is racy in that
1221 * multiple tasks could decide that the same slice can
1222 * not hold a zpool, which is benign. Also choose
1223 * double the number of processors; we hold a lot of
1224 * locks in the kernel, so going beyond this doesn't
1225 * buy us much.
1226 */
1227 t = tpool_create(1, 2 * sysconf(_SC_NPROCESSORS_ONLN),
1228 0, NULL);
1229 for (slice = avl_first(&slice_cache); slice;
1230 (slice = avl_walk(&slice_cache, slice,
1231 AVL_AFTER)))
1232 (void) tpool_dispatch(t, zpool_open_func, slice);
1233 tpool_wait(t);
1234 tpool_destroy(t);
1235
1236 cookie = NULL;
1237 while ((slice = avl_destroy_nodes(&slice_cache,
1238 &cookie)) != NULL) {
1239 if (slice->rn_config != NULL) {
1240 nvlist_t *config = slice->rn_config;
1241 boolean_t matched = B_TRUE;
1242
1243 if (iarg->poolname != NULL) {
1244 char *pname;
1245
1246 matched = nvlist_lookup_string(config,
1247 ZPOOL_CONFIG_POOL_NAME,
1248 &pname) == 0 &&
1249 strcmp(iarg->poolname, pname) == 0;
1250 } else if (iarg->guid != 0) {
1251 uint64_t this_guid;
1252
1253 matched = nvlist_lookup_uint64(config,
1254 ZPOOL_CONFIG_POOL_GUID,
1255 &this_guid) == 0 &&
1256 iarg->guid == this_guid;
1257 }
1258 if (!matched) {
1259 nvlist_free(config);
1260 config = NULL;
1261 continue;
1262 }
1263 /* use the non-raw path for the config */
1264 (void) strlcpy(end, slice->rn_name, pathleft);
1265 if (add_config(hdl, &pools, path, config) != 0)
1266 goto error;
1267 }
1268 free(slice->rn_name);
1269 free(slice);
1270 }
1271 avl_destroy(&slice_cache);
1272
1273 (void) closedir(dirp);
1274 dirp = NULL;
1275 }
1276
1277 ret = get_configs(hdl, &pools, iarg->can_be_active);
1278
1279 error:
1280 for (pe = pools.pools; pe != NULL; pe = penext) {
1281 penext = pe->pe_next;
1282 for (ve = pe->pe_vdevs; ve != NULL; ve = venext) {
1283 venext = ve->ve_next;
1284 for (ce = ve->ve_configs; ce != NULL; ce = cenext) {
1285 cenext = ce->ce_next;
1286 if (ce->ce_config)
1287 nvlist_free(ce->ce_config);
1288 free(ce);
1289 }
1290 free(ve);
1291 }
1292 free(pe);
1293 }
1294
1295 for (ne = pools.names; ne != NULL; ne = nenext) {
1296 nenext = ne->ne_next;
1297 if (ne->ne_name)
1298 free(ne->ne_name);
1299 free(ne);
1300 }
1301
1302 if (dirp)
1303 (void) closedir(dirp);
1304
1305 return (ret);
1306 }
1307
1308 nvlist_t *
1309 zpool_find_import(libzfs_handle_t *hdl, int argc, char **argv)
1310 {
1311 importargs_t iarg = { 0 };
1312
1313 iarg.paths = argc;
1314 iarg.path = argv;
1315
1316 return (zpool_find_import_impl(hdl, &iarg));
1317 }
1318
1319 /*
1320 * Given a cache file, return the contents as a list of importable pools.
1321 * poolname or guid (but not both) are provided by the caller when trying
1322 * to import a specific pool.
1323 */
1324 nvlist_t *
1325 zpool_find_import_cached(libzfs_handle_t *hdl, const char *cachefile,
1326 char *poolname, uint64_t guid)
1327 {
1328 char *buf;
1329 int fd;
1330 struct stat64 statbuf;
1331 nvlist_t *raw, *src, *dst;
1332 nvlist_t *pools;
1333 nvpair_t *elem;
1334 char *name;
1335 uint64_t this_guid;
1336 boolean_t active;
1337
1338 verify(poolname == NULL || guid == 0);
1339
1340 if ((fd = open(cachefile, O_RDONLY)) < 0) {
1341 zfs_error_aux(hdl, "%s", strerror(errno));
1342 (void) zfs_error(hdl, EZFS_BADCACHE,
1343 dgettext(TEXT_DOMAIN, "failed to open cache file"));
1344 return (NULL);
1345 }
1346
1347 if (fstat64(fd, &statbuf) != 0) {
1348 zfs_error_aux(hdl, "%s", strerror(errno));
1349 (void) close(fd);
1350 (void) zfs_error(hdl, EZFS_BADCACHE,
1351 dgettext(TEXT_DOMAIN, "failed to get size of cache file"));
1352 return (NULL);
1353 }
1354
1355 if ((buf = zfs_alloc(hdl, statbuf.st_size)) == NULL) {
1356 (void) close(fd);
1357 return (NULL);
1358 }
1359
1360 if (read(fd, buf, statbuf.st_size) != statbuf.st_size) {
1361 (void) close(fd);
1362 free(buf);
1363 (void) zfs_error(hdl, EZFS_BADCACHE,
1364 dgettext(TEXT_DOMAIN,
1365 "failed to read cache file contents"));
1366 return (NULL);
1367 }
1368
1369 (void) close(fd);
1370
1371 if (nvlist_unpack(buf, statbuf.st_size, &raw, 0) != 0) {
1372 free(buf);
1373 (void) zfs_error(hdl, EZFS_BADCACHE,
1374 dgettext(TEXT_DOMAIN,
1375 "invalid or corrupt cache file contents"));
1376 return (NULL);
1377 }
1378
1379 free(buf);
1380
1381 /*
1382 * Go through and get the current state of the pools and refresh their
1383 * state.
1384 */
1385 if (nvlist_alloc(&pools, 0, 0) != 0) {
1386 (void) no_memory(hdl);
1387 nvlist_free(raw);
1388 return (NULL);
1389 }
1390
1391 elem = NULL;
1392 while ((elem = nvlist_next_nvpair(raw, elem)) != NULL) {
1393 verify(nvpair_value_nvlist(elem, &src) == 0);
1394
1395 verify(nvlist_lookup_string(src, ZPOOL_CONFIG_POOL_NAME,
1396 &name) == 0);
1397 if (poolname != NULL && strcmp(poolname, name) != 0)
1398 continue;
1399
1400 verify(nvlist_lookup_uint64(src, ZPOOL_CONFIG_POOL_GUID,
1401 &this_guid) == 0);
1402 if (guid != 0) {
1403 verify(nvlist_lookup_uint64(src, ZPOOL_CONFIG_POOL_GUID,
1404 &this_guid) == 0);
1405 if (guid != this_guid)
1406 continue;
1407 }
1408
1409 if (pool_active(hdl, name, this_guid, &active) != 0) {
1410 nvlist_free(raw);
1411 nvlist_free(pools);
1412 return (NULL);
1413 }
1414
1415 if (active)
1416 continue;
1417
1418 if ((dst = refresh_config(hdl, src)) == NULL) {
1419 nvlist_free(raw);
1420 nvlist_free(pools);
1421 return (NULL);
1422 }
1423
1424 if (nvlist_add_nvlist(pools, nvpair_name(elem), dst) != 0) {
1425 (void) no_memory(hdl);
1426 nvlist_free(dst);
1427 nvlist_free(raw);
1428 nvlist_free(pools);
1429 return (NULL);
1430 }
1431 nvlist_free(dst);
1432 }
1433
1434 nvlist_free(raw);
1435 return (pools);
1436 }
1437
1438 static int
1439 name_or_guid_exists(zpool_handle_t *zhp, void *data)
1440 {
1441 importargs_t *import = data;
1442 int found = 0;
1443
1444 if (import->poolname != NULL) {
1445 char *pool_name;
1446
1447 verify(nvlist_lookup_string(zhp->zpool_config,
1448 ZPOOL_CONFIG_POOL_NAME, &pool_name) == 0);
1449 if (strcmp(pool_name, import->poolname) == 0)
1450 found = 1;
1451 } else {
1452 uint64_t pool_guid;
1453
1454 verify(nvlist_lookup_uint64(zhp->zpool_config,
1455 ZPOOL_CONFIG_POOL_GUID, &pool_guid) == 0);
1456 if (pool_guid == import->guid)
1457 found = 1;
1458 }
1459
1460 zpool_close(zhp);
1461 return (found);
1462 }
1463
1464 nvlist_t *
1465 zpool_search_import(libzfs_handle_t *hdl, importargs_t *import)
1466 {
1467 verify(import->poolname == NULL || import->guid == 0);
1468
1469 if (import->unique)
1470 import->exists = zpool_iter(hdl, name_or_guid_exists, import);
1471
1472 if (import->cachefile != NULL)
1473 return (zpool_find_import_cached(hdl, import->cachefile,
1474 import->poolname, import->guid));
1475
1476 return (zpool_find_import_impl(hdl, import));
1477 }
1478
1479 boolean_t
1480 find_guid(nvlist_t *nv, uint64_t guid)
1481 {
1482 uint64_t tmp;
1483 nvlist_t **child;
1484 uint_t c, children;
1485
1486 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &tmp) == 0);
1487 if (tmp == guid)
1488 return (B_TRUE);
1489
1490 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
1491 &child, &children) == 0) {
1492 for (c = 0; c < children; c++)
1493 if (find_guid(child[c], guid))
1494 return (B_TRUE);
1495 }
1496
1497 return (B_FALSE);
1498 }
1499
1500 typedef struct aux_cbdata {
1501 const char *cb_type;
1502 uint64_t cb_guid;
1503 zpool_handle_t *cb_zhp;
1504 } aux_cbdata_t;
1505
1506 static int
1507 find_aux(zpool_handle_t *zhp, void *data)
1508 {
1509 aux_cbdata_t *cbp = data;
1510 nvlist_t **list;
1511 uint_t i, count;
1512 uint64_t guid;
1513 nvlist_t *nvroot;
1514
1515 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE,
1516 &nvroot) == 0);
1517
1518 if (nvlist_lookup_nvlist_array(nvroot, cbp->cb_type,
1519 &list, &count) == 0) {
1520 for (i = 0; i < count; i++) {
1521 verify(nvlist_lookup_uint64(list[i],
1522 ZPOOL_CONFIG_GUID, &guid) == 0);
1523 if (guid == cbp->cb_guid) {
1524 cbp->cb_zhp = zhp;
1525 return (1);
1526 }
1527 }
1528 }
1529
1530 zpool_close(zhp);
1531 return (0);
1532 }
1533
1534 /*
1535 * Determines if the pool is in use. If so, it returns true and the state of
1536 * the pool as well as the name of the pool. Both strings are allocated and
1537 * must be freed by the caller.
1538 */
1539 int
1540 zpool_in_use(libzfs_handle_t *hdl, int fd, pool_state_t *state, char **namestr,
1541 boolean_t *inuse)
1542 {
1543 nvlist_t *config;
1544 char *name;
1545 boolean_t ret;
1546 uint64_t guid, vdev_guid;
1547 zpool_handle_t *zhp;
1548 nvlist_t *pool_config;
1549 uint64_t stateval, isspare;
1550 aux_cbdata_t cb = { 0 };
1551 boolean_t isactive;
1552
1553 *inuse = B_FALSE;
1554
1555 if (zpool_read_label(fd, &config) != 0) {
1556 (void) no_memory(hdl);
1557 return (-1);
1558 }
1559
1560 if (config == NULL)
1561 return (0);
1562
1563 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE,
1564 &stateval) == 0);
1565 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID,
1566 &vdev_guid) == 0);
1567
1568 if (stateval != POOL_STATE_SPARE && stateval != POOL_STATE_L2CACHE) {
1569 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
1570 &name) == 0);
1571 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
1572 &guid) == 0);
1573 }
1574
1575 switch (stateval) {
1576 case POOL_STATE_EXPORTED:
1577 /*
1578 * A pool with an exported state may in fact be imported
1579 * read-only, so check the in-core state to see if it's
1580 * active and imported read-only. If it is, set
1581 * its state to active.
1582 */
1583 if (pool_active(hdl, name, guid, &isactive) == 0 && isactive &&
1584 (zhp = zpool_open_canfail(hdl, name)) != NULL &&
1585 zpool_get_prop_int(zhp, ZPOOL_PROP_READONLY, NULL))
1586 stateval = POOL_STATE_ACTIVE;
1587
1588 ret = B_TRUE;
1589 break;
1590
1591 case POOL_STATE_ACTIVE:
1592 /*
1593 * For an active pool, we have to determine if it's really part
1594 * of a currently active pool (in which case the pool will exist
1595 * and the guid will be the same), or whether it's part of an
1596 * active pool that was disconnected without being explicitly
1597 * exported.
1598 */
1599 if (pool_active(hdl, name, guid, &isactive) != 0) {
1600 nvlist_free(config);
1601 return (-1);
1602 }
1603
1604 if (isactive) {
1605 /*
1606 * Because the device may have been removed while
1607 * offlined, we only report it as active if the vdev is
1608 * still present in the config. Otherwise, pretend like
1609 * it's not in use.
1610 */
1611 if ((zhp = zpool_open_canfail(hdl, name)) != NULL &&
1612 (pool_config = zpool_get_config(zhp, NULL))
1613 != NULL) {
1614 nvlist_t *nvroot;
1615
1616 verify(nvlist_lookup_nvlist(pool_config,
1617 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
1618 ret = find_guid(nvroot, vdev_guid);
1619 } else {
1620 ret = B_FALSE;
1621 }
1622
1623 /*
1624 * If this is an active spare within another pool, we
1625 * treat it like an unused hot spare. This allows the
1626 * user to create a pool with a hot spare that currently
1627 * in use within another pool. Since we return B_TRUE,
1628 * libdiskmgt will continue to prevent generic consumers
1629 * from using the device.
1630 */
1631 if (ret && nvlist_lookup_uint64(config,
1632 ZPOOL_CONFIG_IS_SPARE, &isspare) == 0 && isspare)
1633 stateval = POOL_STATE_SPARE;
1634
1635 if (zhp != NULL)
1636 zpool_close(zhp);
1637 } else {
1638 stateval = POOL_STATE_POTENTIALLY_ACTIVE;
1639 ret = B_TRUE;
1640 }
1641 break;
1642
1643 case POOL_STATE_SPARE:
1644 /*
1645 * For a hot spare, it can be either definitively in use, or
1646 * potentially active. To determine if it's in use, we iterate
1647 * over all pools in the system and search for one with a spare
1648 * with a matching guid.
1649 *
1650 * Due to the shared nature of spares, we don't actually report
1651 * the potentially active case as in use. This means the user
1652 * can freely create pools on the hot spares of exported pools,
1653 * but to do otherwise makes the resulting code complicated, and
1654 * we end up having to deal with this case anyway.
1655 */
1656 cb.cb_zhp = NULL;
1657 cb.cb_guid = vdev_guid;
1658 cb.cb_type = ZPOOL_CONFIG_SPARES;
1659 if (zpool_iter(hdl, find_aux, &cb) == 1) {
1660 name = (char *)zpool_get_name(cb.cb_zhp);
1661 ret = TRUE;
1662 } else {
1663 ret = FALSE;
1664 }
1665 break;
1666
1667 case POOL_STATE_L2CACHE:
1668
1669 /*
1670 * Check if any pool is currently using this l2cache device.
1671 */
1672 cb.cb_zhp = NULL;
1673 cb.cb_guid = vdev_guid;
1674 cb.cb_type = ZPOOL_CONFIG_L2CACHE;
1675 if (zpool_iter(hdl, find_aux, &cb) == 1) {
1676 name = (char *)zpool_get_name(cb.cb_zhp);
1677 ret = TRUE;
1678 } else {
1679 ret = FALSE;
1680 }
1681 break;
1682
1683 default:
1684 ret = B_FALSE;
1685 }
1686
1687
1688 if (ret) {
1689 if ((*namestr = zfs_strdup(hdl, name)) == NULL) {
1690 if (cb.cb_zhp)
1691 zpool_close(cb.cb_zhp);
1692 nvlist_free(config);
1693 return (-1);
1694 }
1695 *state = (pool_state_t)stateval;
1696 }
1697
1698 if (cb.cb_zhp)
1699 zpool_close(cb.cb_zhp);
1700
1701 nvlist_free(config);
1702 *inuse = ret;
1703 return (0);
1704 }